Copilot in agent mode is powerful. But one AI doing everything isn't a system.I kept rewriting the same instructions every session. Context was lost.
Results were inconsistent. So I built OMG (Oh-My-GithubCopilot).
What it adds on top of Copilot:
- 28 specialized agents (code review, security, TDD, architecture...)
- 22 reusable skills — define once, use anywhere
- MCP-based memory that actually persists across sessions
- Safety hooks for autonomous runs
New in v1.1.6 — Hook system improvement, ECC integration:
- Five core skills now fire structured multiple-choice prompts at decision gates, mirroring OMC's gateway-level interrupt hooks
- Language-specialist reviewer agents
- TDD-first by default
-
omg-autopilotfor end-to-end autonomous builds -
ralphloop — runs until the task is done, not until it gives up
**One line summary:
Copilot gives speed. OMG gives system.**
🐙 GitHub: https://github.com/jmstar85/oh-my-githubcopilot
🔗 Marketplace: https://marketplace.visualstudio.com/items?itemName=jmstar85.oh-my-githubcopilot&ssr=false#overview
Happy to answer questions — especially on how agent orchestration works in practice.
Top comments (0)